Team:SYSU-CHINA/Design

design
Overview
In our project, an algorithm-guided model is trained from natural substrates of ADAR1 and used to establish a candidate dsRNA library. Those candidates are connected with sequences to help them cyclize and transfer into HeLa cells. In cells, they compete to bind ADAR1 with an editable stem-loop which has a toxic gene downstream of it. The whole system is regulated by IFN-α and Tet-on system. The cells will only survive when endogenous ADAR1 is inhibited by transferred dsRNAs. Efficient substrates are extracted and used to train the model above for the next round. Through the continuous cycle of this screening process, we can obtain high-affinity inhibitors of ADAR1 efficiently, and extend this model to other RBPs.
Figure1.Flow chart of the project
In 2020, Song, Y. et al.[1] published a research showing that the improved technology irCLASH can obtain the sequence and quantity information of dsRNA as ADAR substrates comprehensively and effectively. At the same time, the research revealed the possible characteristics of some ADAR substrates. We obtained 4894 exogenous ADAR1 substrates and 8558 endogenous sequences of ADAR1 substrates, as well as chromosome position, gene position, affinity and other information. These thousands of sequences make up our original dsRNA library.
Experimental Design
The main idea of our experimental design is to combine ADAR1 editing event with a trait easy to be selected, then adjust the level of selected pressure to gradually obtain inhibitors with high affinity to ADAR1. To put it simply, it is a question of competing to bind with ADAR1, consisting of four parts: The competitors, the intensity of competition, the place to compete and the introduction of fresh blood.
The competitors
The dsRNAs and the stem-loop are forced to compete binding with ADAR1 in cells. When the former ones have higher affinity, the cell survives; Conversely, the cell dies.
  • Cycled dsRNA
  • In order to improve the intracellular stability of dsRNA, we add ligation sequence and ribozyme sequence at both ends of it. Once the transcript expresses, it first undergoes self-shear under the action of ribozyme sequence, and then cyclizes through the action of RtcB, which is very common in cells.
    Figure 2, 3. Electrophoretogram showing self-cleaved and cyclization of sequences in vitro[2]
  • Stem-loop and toxic gene
  • Stem-loop is used to control the translation of toxic gene. It utilizes the natural GluA2 editing substrate in which the R/G editing site has been modified into a stop codon (UAG) that upon editing is recoded into a tryptophan codon (UGG). In the literature we refer to, it is designed to construct bioluminescence reporting systems that can rapidly and accurately measure endogenous editing activity. A firefly luciferase reporter gene upstream of the edited site monitors translation, and a nanoluciferase reporter gene downstream is used to measure read-through after editing. The editing efficiency of ADAR is reflected by fluorescence ratio.
    Figure 4. The stem-loop modified from Natural GluA2 editing Substrate[3]
    Here, we place a toxic gene downstream of the stem-loop. When the affinity of dsRNA is insufficient, the stem-loop will bind with ADAR1 and be edited, so the toxic gene will express, resulting in the death of cells as well as the elimination of the corresponding dsRNA. We also assembled the GFP gene upstream of the stem-loop for expression detection. The effect of the stem ring is simulated by comparing the experimental group with the positive control (simulated 100% coating, amber termination in stem loop: A→G) and the negative control (simulated 0% coating,18nt miss at the edited part of the stem-loop).
    Figure 5. Three different groups of stem-loop
    As for the toxic gene, since one of the potential applications of our project is to help develop cancer drugs, we paid more attention on those that acts in cancer cells and believed to have less effects on normal cells. In 2013, iGEM team SYSU-China has rigorously characterized several potentially useful toxic genes [4]. After searching for more literature and comparing, we finally listed Apoptin, which has been proved to be safe for many normal cell types, as our first choice. If conditions permit, we will also check more different toxic genes to explore the most effective one.
    The intensity of competition
    In order to better control the intensity of competition, we use IFN-α to adjust ADAR1 editing level. At the same time, we introduce Tet-on system to regulate the expression of stem loop - toxic gene components. For the candidate dsRNA, IFN-α adjusts the amount of ADAR1 it needs to bind with, while Tet-on system changes the number of its competitors. In this way, the selection pressure of dsRNA can be effectively regulated.
  • IFN-α
  • The literature shows that IFN-α can induce the editing level of ADAR1. With the increase of IFN-α concentration, the editing level of ADAR1 also increases, and there is a correlation between the two. Therefore, we take it as one of the means to adjust the selection pressure.
    Figure 6. Induction of A-to-I editing by IFN-α in HeLa cells[3]
  • Dox
  • In order to prevent the stem ring from being edited before dsRNA accumulates to an effective concentration, which leads to cell death and thus the wrong selection of dsRNA, we hope to control the expression of the stem-loop through exogenous compounds. Tet control system is an ideal choice, and since our goal is that the transcriptome expresses only when we need it, we chose the Tet-on instead of Tet-off system.
    Figure 7. The function mechanism of tet-on system
    Another advantage of the introduction of Tet-on system is that we can control the transcriptional level of stem-loop by adjusting the Dox concentration, which broadens the adjustable range of selective pressure.
    The place to compete
    In addition to component construction, it is also important to select appropriate chassis. Initially, our list of candidates included yeast, HEK293 cells and Hela cells, three different types of cells that are commonly used in biological research. We also found the introduction of dual luciferase reporting system with stem-loop in yeast, so in theory it can function in all three types of cells. After searching for more information and discussions, we finally chose to conduct the experiment in Hela cells.
  • Why not yeast:
  • 1. Yeast does not express ADAR1 itself and relevant studies rely on exogenous introduction. And the editing level of exogenous ADAR1 is poor

    2. Technique of dsRNA cyclization relies on endogenous nucleic acid ligase in mammalian cell

    3. Considering the purposes of our project, experiments carried out with yeast may be lack of clinical significance

    Figure 8. Editing level of ADAR1 in yeast[5]
  • Why not HEK293:
  • 1. The editing level of endogenous ADAR1 in HEK293 is low

    2. Hela is widely used in research of cancer so it may be more related to our project

    Figure 9. The transfected reporter sensing endogenous editing in different human cell lines[3]
    The introduction of fresh blood
    Obviously, for directed evolution, it is not enough to work on only one sequence, so we need to update the dsRNA library in a convenient and effective way as possible and constantly introduce competitors with higher affinity to ADAR1.To achieve this, on the one hand, we design an algorithm to extract and predict the sequence characteristics of the high-affinity substrates, and rationally construct the dsRNA library under the guidance of the algorithm. On the other hand, we introduce random mutations to the original dsRNA sequence through error-prone PCR, enrich the dsRNA library and provide different selection results for the algorithm to train the model for the next round.
    Figure 10. Introduction of new sequences and cyclic selection
    Algorithm Design

    In order to speed up the progress of directed evolution and reduce the experimental quantity of this huge work, we adopt the experimental method combining semi-rational design and directed evolution. Directed evolution has emerged in experiments, in which specific selection pressures are applied to a population that already has a mutant sequence, forcing the population to evolve in a way that adapts to that pressure. Semi-rational design, however, refers to the use of feature extraction and machine learning algorithms to predict the sequence of dsRNAs with high affinity to ADAR1.

    First, we extract the characteristics of dsRNAs for analysis, and the simplest one should be the first-order sequence of dsRNAs.

    Furthermore, we innovatively introduce the relevant knowledge of machine learning field into the project. Based on the corresponding situation of dsRNA and its expression effect obtained in previous experiments, the data set is constructed and used as the training set. In this way, the neural network model can be used to predict the effect of training on the basis of the neural network.

    Finally, the computer can predict and judge the expression effect in advance by inputting RNA, so as to reduce the workload of experimenters.

    For more information, please click here: https://2020.igem.org/Team:SYSU-CHINA/Model

    Reference

    [1] Song, Y., Yang, W., Fu, Q. et al. irCLASH reveals RNA substrates recognized by human ADARs. Nat Struct Mol Biol 27, 351–362 (2020). https://doi.org/10.1038/s41594-020-0398-4

    [2] Litke JL, Jaffrey SR. Highly efficient expression of circular RNA aptamers in cells using autocatalytic transcripts. Nat Biotechnol. 2019 Jun;37(6):667-675. doi: 10.1038/s41587-019-0090-6. Epub 2019 Apr 8. PMID: 30962542; PMCID: PMC6554452.

    [3] Fritzell K, Xu LD, Otrocka M, Andréasson C, Öhman M. Sensitive ADAR editing reporter in cancer cells enables high-throughput screening of small molecule libraries. Nucleic Acids Res. 2019 Feb 28;47(4):e22. doi: 10.1093/nar/gky1228. PMID: 30590609; PMCID: PMC6393238.

    [4] https://2013.igem.org/Team:SYSU-China/Project/Design

    [5] Garncarz W, Tariq A, Handl C, et al. A high-throughput screen to identify enhancers of ADAR-mediated RNA-editing[J]. RNA Biology, 2013, 10(2): 192-204.